feat(registry): workspace hibernation — auto-pause idle workspaces

platform/cmd/server/main.go

@@ -185,6 +185,13 @@ func main() {
 	cronSched := scheduler.New(wh, broadcaster)
 	go supervised.RunWithRecover(ctx, "scheduler", cronSched.Start)
 
+	// Hibernation Monitor — auto-pauses idle workspaces that have
+	// hibernation_idle_minutes configured (#711). Wakeup is triggered
+	// automatically on the next incoming A2A message.
+	go supervised.RunWithRecover(ctx, "hibernation-monitor", func(c context.Context) {
+		registry.StartHibernationMonitor(c, wh.HibernateWorkspace)
+	})
+
 	// Channel Manager — social channel integrations (Telegram, Slack, etc.)
 	channelMgr := channels.NewManager(wh, broadcaster)
 	go supervised.RunWithRecover(ctx, "channel-manager", channelMgr.Start)

platform/internal/handlers/a2a_proxy.go

@@ -274,16 +274,16 @@ func (h *WorkspaceHandler) proxyA2ARequest(ctx context.Context, workspaceID stri
 	}
 	defer resp.Body.Close()
 
-	// Read agent response (capped at 10MB)
+	// Read agent response (capped at 10MB).
+	// #689: Do() succeeded, which means the target received the request and sent
+	// back response headers — delivery is confirmed. The body couldn't be
+	// fully read (connection drop, timeout mid-stream). Surface
+	// delivery_confirmed so callers can distinguish "not delivered" from
+	// "delivered, but response body lost". When delivery is confirmed,
+	// log the activity as successful (delivery happened) rather than leaving
+	// a false "failed" entry in the audit trail.
 	respBody, readErr := io.ReadAll(io.LimitReader(resp.Body, maxProxyResponseBody))
 	if readErr != nil {
-		// Do() succeeded, which means the target received the request and sent
-		// back response headers — delivery is confirmed. The body couldn't be
-		// fully read (connection drop, timeout mid-stream). Surface
-		// delivery_confirmed so callers can distinguish "not delivered" from
-		// "delivered, but response body lost" (#689). When delivery is confirmed,
-		// log the activity as successful (delivery happened) rather than leaving
-		// a false "failed" entry in the audit trail.
 		deliveryConfirmed := resp.StatusCode >= 200 && resp.StatusCode < 400
 		log.Printf("ProxyA2A: body read failed for %s (status=%d delivery_confirmed=%v bytes_read=%d): %v",
 			workspaceID, resp.StatusCode, deliveryConfirmed, len(respBody), readErr)
@@ -338,6 +338,22 @@ func (h *WorkspaceHandler) resolveAgentURL(ctx context.Context, workspaceID stri
 			}
 		}
 		if !urlNullable.Valid || urlNullable.String == "" {
+			// Auto-wake hibernated workspace on incoming A2A message (#711).
+			// Re-provision asynchronously and return 503 with a retry hint so
+			// the caller can retry once the workspace is back online (~10s).
+			if status == "hibernated" {
+				log.Printf("ProxyA2A: waking hibernated workspace %s", workspaceID)
+				go h.RestartByID(workspaceID)
+				return "", &proxyA2AError{
+					Status:  http.StatusServiceUnavailable,
+					Headers: map[string]string{"Retry-After": "15"},
+					Response: gin.H{
+						"error":       "workspace is waking from hibernation — retry in ~15 seconds",
+						"waking":      true,
+						"retry_after": 15,
+					},
+				}
+			}
 			return "", &proxyA2AError{
 				Status:   http.StatusServiceUnavailable,
 				Response: gin.H{"error": "workspace has no URL", "status": status},

platform/internal/handlers/a2a_proxy_test.go

@@ -1237,3 +1237,81 @@ func TestLogA2ASuccess_ErrorStatus(t *testing.T) {
 	handler.logA2ASuccess(context.Background(), "ws-err", "ws-caller", []byte(`{}`), []byte(`{}`), "message/send", 500, 10)
 	time.Sleep(80 * time.Millisecond)
 }
+
+// ──────────────────────────────────────────────────────────────────────────────
+// A2A auto-wake: hibernated workspace (#711)
+// ──────────────────────────────────────────────────────────────────────────────
+
+// TestResolveAgentURL_HibernatedWorkspace_Returns503WithWaking verifies the
+// auto-wake path added in PR #724: when resolveAgentURL finds a workspace with
+// status='hibernated' and no URL, it must:
+//   - Return a proxyA2AError with Status 503
+//   - Set Retry-After: 15 in Headers
+//   - Include waking:true and retry_after:15 in the response body
+//
+// RestartByID fires asynchronously via `go h.RestartByID(workspaceID)`. Because
+// provisioner is nil in tests, RestartByID returns immediately without any DB
+// calls, so no additional mocks are needed.
+func TestResolveAgentURL_HibernatedWorkspace_Returns503WithWaking(t *testing.T) {
+	mock := setupTestDB(t)
+	setupTestRedis(t) // empty Redis → GetCachedURL returns error → DB fallback
+
+	handler := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())
+
+	// DB fallback: workspace exists but has no URL and is hibernated.
+	mock.ExpectQuery(`SELECT url, status FROM workspaces WHERE id =`).
+		WithArgs("ws-hibernated").
+		WillReturnRows(sqlmock.NewRows([]string{"url", "status"}).AddRow("", "hibernated"))
+
+	_, perr := handler.resolveAgentURL(context.Background(), "ws-hibernated")
+
+	if perr == nil {
+		t.Fatal("expected proxyA2AError, got nil")
+	}
+	if perr.Status != http.StatusServiceUnavailable {
+		t.Errorf("expected status 503, got %d", perr.Status)
+	}
+	if perr.Headers["Retry-After"] != "15" {
+		t.Errorf("expected Retry-After: 15, got %q", perr.Headers["Retry-After"])
+	}
+
+	if perr.Response["waking"] != true {
+		t.Errorf("expected waking:true in body, got %v", perr.Response["waking"])
+	}
+	if perr.Response["retry_after"] != 15 {
+		t.Errorf("expected retry_after:15 in body, got %v", perr.Response["retry_after"])
+	}
+
+	if err := mock.ExpectationsWereMet(); err != nil {
+		t.Errorf("unmet DB expectations: %v", err)
+	}
+}
+
+// TestResolveAgentURL_HibernatedWorkspace_NullURLVariant verifies the same
+// auto-wake behaviour when the DB returns a SQL NULL for the url column
+// (rather than an empty string). Both forms represent "no URL assigned".
+func TestResolveAgentURL_HibernatedWorkspace_NullURLVariant(t *testing.T) {
+	mock := setupTestDB(t)
+	setupTestRedis(t)
+	handler := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())
+
+	mock.ExpectQuery(`SELECT url, status FROM workspaces WHERE id =`).
+		WithArgs("ws-hibernated-null").
+		WillReturnRows(sqlmock.NewRows([]string{"url", "status"}).AddRow(nil, "hibernated"))
+
+	_, perr := handler.resolveAgentURL(context.Background(), "ws-hibernated-null")
+
+	if perr == nil {
+		t.Fatal("expected proxyA2AError, got nil")
+	}
+	if perr.Status != http.StatusServiceUnavailable {
+		t.Errorf("expected status 503, got %d", perr.Status)
+	}
+	if perr.Headers["Retry-After"] != "15" {
+		t.Errorf("expected Retry-After: 15, got %q", perr.Headers["Retry-After"])
+	}
+
+	if err := mock.ExpectationsWereMet(); err != nil {
+		t.Errorf("unmet DB expectations: %v", err)
+	}
+}